Reference Design
NVIDIA DGX Spark
NVIDIA · Model: DGX Spark
The original personal AI supercomputer from NVIDIA. Powered by the GB10 Grace Blackwell Superchip with 128GB unified memory and the complete NVIDIA AI software stack.
Contact for Pricing
Warranty: N/A
Inquire About This ProductFull Specifications
| Architecture | NVIDIA Grace Blackwell |
| GPU | NVIDIA Blackwell Architecture |
| CPU | 20-core Arm (10 Cortex-X925 + 10 Cortex-A725) |
| Tensor Performance | 1,000 AI TOPS |
| System Memory | 128GB LPDDR5x, unified system memory |
| Memory Interface | 256-bit |
| Memory Bandwidth | 273 GB/s |
| OS | NVIDIA DGX OS |
| Storage | N/A (Reference Design) |
| USB | 4x USB Type-C |
| Ethernet | 10 GbE, 1x RJ-45 |
| NIC | ConnectX-7 Smart NIC |
| Wi-Fi | Wi-Fi 7 |
| Bluetooth | BT 5.3 |
| Audio Output | HDMI multichannel audio output |
| Power Consumption | 170W |
| Display Connectors | HDMI 2.1a |
| NVENC / NVDEC | 1x / 1x |
| Dimensions | 150 × 150 × 50.5 mm |
| Weight | 1.2 kg |
Other Products
NVIDIA DGX Spark FAQ
What makes the NVIDIA DGX Spark the reference design?+
The NVIDIA DGX Spark is the original design from NVIDIA that all partner variants are based on. It defines the baseline specifications — GB10 Grace Blackwell Superchip, 128GB unified memory, and the complete NVIDIA AI software stack — that other manufacturers then customize.
Why is the NVIDIA DGX Spark listed as 'Contact for Pricing'?+
The NVIDIA DGX Spark reference design is available through NVIDIA's direct sales channel. Pricing and availability depend on NVIDIA's current allocation. Contact us and we'll help you get a quote or recommend an equivalent partner model.
How does the NVIDIA DGX Spark differ from partner models?+
The core hardware is identical across all models — the same GB10 superchip and 128GB memory. Partner models like Gigabyte, MSI, ASUS, Dell, and Leadtek add their own storage configurations, warranty programs, cooling designs, and connectivity options.
What storage does the NVIDIA DGX Spark reference design include?+
The reference design specifications list storage as N/A because NVIDIA allows partner manufacturers to configure storage options. If you need a specific storage capacity, we recommend partner models which offer 1TB or 4TB NVMe options.
Can I use DGX Spark for commercial AI deployment?+
Yes. DGX Spark is designed for both development and production inference workloads. It can serve AI models locally for commercial applications, run fine-tuning jobs, and handle real-time inference — all without cloud dependency.